cartofi altiplano.2

Estimating the potential impact of frost resistant potatocultivars in the Altiplano (Peru and Bolivia)

R. J. Hijmans1, B. Condori2, R. Carrillo2, and M. J. Kropff 3

Abstract

Frost damage causes significant losses in potato production in the Altiplano (Peru and Bolivia). Wequantified the extent to which frost resistant cultivars could alleviate this problem, using a quantitativeand constraint-specific agroecological zonation approach. The LINTUL potato growth simulation modelwas adapted to incorporate the effect of frost damage on yield, and calibrated using experimental data.High-resolution grids of monthly climate data were created for a number of variables, includingabsolute minimum temperature and its standard deviation, and used as input for the simulation model.The model was run for each grid cell, using a standard potato cultivar in which frost resistanceparameters were changed in increments of 1C. A geo-referenced database of potato distribution wasused to process the output of the simulation model to calculate potato-area weighted results. The resultsindicate that when frost resistance increases from 1C (current level) to 2C or 3C, average potatoyield would increase 26 and 40%, respectively. After that, the effect flattens off and a further increase inresistance leads to only a small increase in simulated potato yield.

1 International Potato Center (CIP), P.O. Box 1558, Lima 12, Peru. E-mail: [email protected] Fundacin para la Promocin e Investigacin de Productos Andinos (PROINPA), La Paz, Bolivia.3 Chairgroup Crop and Weed Ecology, Department of Plant Sciences, Wageningen University, Wageningen, The

Netherlands.

INTRODUCTION

The Altiplano is a high plateau in the Andes ofPeru and Bolivia. In this study we focus on the part ofthe Altiplano called the TDPS system, named afterthe catchment of Lake Titicaca, the DesaguaderoRiver, Lake Poop, and the Salt Lake of Coipasa(OEA, 1996) (Figure 1). That excludes the southernmostpart of the Altiplano, which is rather arid, sparselypopulated, and less important for agriculture.Seventy-five percent of the TDPS system, hereinaftercalled the Altiplano, is between 3,600 and 4,300 m.above sea level; the other 25% is higher. It comprises149,000 km2. Lake Titicaca, covering 8,400 km2, is aconspicuous part of the topography that greatlyinfluences local precipitation and temperature.

In 1993, the Altiplano had about 2.2 millioninhabitants (OEA, 1996). The Altiplano is one of thepoorest areas of the Americas, and poorer than mostother parts of Bolivia and Peru. About 65% of theeconomically active population is engaged inagriculture (OEA, 1996). Most of the cropland islocated below 4000 m; above that elevation land isused for grazing. Potato is by far the economicallymost important crop, accounting for 63% of the grossvalue of crop production (OEA, 1996). The areaplanted to potato is about 63,000 ha (G-DRU, 1996;INEI, 1996). Reported potato yields are low, at 5.2t/ha in the Peruvian and northern Bolivian sections ofthe Altiplano, and 3.6 t/ha in the southern part of theBolivian section (OEA, 1996; G-DRU, 1996).However, an extensive survey in four Departments ofBolivia (outside of the Altiplano) indicates thatgovernment statistics have underreported yields by asmuch as 50% (Terrazas et al., 1998).

Proceedings - The Third International Symposium on Systems Approaches for Agricultural Development

2

Figure 1. The Altiplano (TDPS system) in the Andes of Peru and Bolivia. Topography and distribution of potato production and thelocation of weather stations used for the construction of the climate surfaces.

The growing season in the Altiplano extendsbetween October and March, when maximum annualtemperature coincides with the rainy season. In theagricultural zones, maximum temperature is around18C and minimum temperature around 4C duringthe growing season (INTECSA, 1993; Frre et al.,1975). Precipitation is highest in the northeast and inperipheral areas of Lake Titicaca, around 800mm/year, and lowest, about 200 mm, in thesouthwest. Production risk for potato is high due to avariety of factors, particularly drought, hail, and frost.Although there is an average frost-free period ofabout 140 d for the northern Altiplano and 110 d for

the southern Altiplano (Le Tacon, 1989), night frostscaused by radiative cooling on clear nights mayoccur at any time during the growing season. Themid-season frost problem is common throughout thetropical highlands: see, e.g., Knapp (1988), for adescription of frost incidence and potato productionin Ecuador. However this is unlike potato productionconditions in temperate regions, where frosts occuronly at the beginning and end of the growing season.Booy (1961) describes a case of night frost damageon potatoes in the beginning of the growing seasonin the Netherlands.

Estimating the Potential Impact of Frost Resistant Potato Cultivars in the Altiplano (Peru and Bolivia)

3

Frost occurs when air temperature near the Earthssurface drops below 0C (Kalma et al., 1992). Whenfrost damage is described in the literature, it is notalways clear whether the reported temperatures referto conditions at screen height (1.5 to 2 m) or at cropcanopy height. In this study, temperature refers toscreen height conditions. During a frost event, airtemperature may be as much as 1C lower at potatocanopy height (De Bouet du Portal, 1993), andleaves might be colder still. The temperature atwhich frost damage occurs depends on the speciesand the cultivar. For S. tuberosum subsp. andigena,the most commonly grown potato in the Andes, frostdamage is likely to occur when the temperaturedrops to -2C or lower (Carrasco et al. 1997). Higherfrost resistance exists in other cultivated and wildpotato species. For example, cultivated potatospecies such as S. ajanhuiri and S. curtilobum incurdamage at -3 to -5C (Huanco, 1992; Tapia andSaravia, 1997), whereas S. juzepzuckii generallyresists temperatures down to -5C and lower (Huanco1992, Canahua and Aguilar 1992, Tapia and Saravia1997). With the exception of S. ajanhuiri, the tubersof the species with higher frost resistance tend to bebitter due to high levels of glycoalkaloids andtherefore require processing before consumption.Henceforward, non-tuberosum cultivated potatospecies will be referred to as "bitter potatoes".

Hijmans (1999) estimated that 25% of total areaplanted to potato in the Altiplano has an extremelyhigh (> 33 %) frost risk of a -2C event occurringonce every 3 yr. He concluded that this high-riskarea is most likely planted with bitter potatoes. Thiswas supported by Canahua and Aguilar (1992) andHuanco (1992), who estimate that about one-third oftotal potato area of the Peruvian Altiplano is plantedwith bitter potatoes, of which 60% are S.juzepzuckii, and 33% are S. curtilobum (Canahuaand Aguilar 1992). It was also supported by Reas(1992) estimate that bitter potatoes comprise 15% oftotal potato area in Bolivia, where more bitterpotatoes are found in the Altiplano than in mostother zones.

Frost can cause partial or complete loss of leafarea of a potato crop, leading to a reduction inphotosynthesis and hence yield. In turn, crop failurecaused by frost damage may lead to a decrease in thetotal area planted to potato in the subsequent seasondue to seed shortage (Morlon 1989). The highproduction risks presented by frost and other factorsmay also lead to less investment in agriculture,resulting in decreased production, despite of theweather conditions in a given year.

Farmers can prevent or reduce frost damage byplanting potatoes on warm soil (with a high thermalconductivity, (cf. Booy 1961)) and on slopes, wherefrost incidence is lower than on the valley floor (DeBouet du Portal 1993); applying frost-relatedmanagement practices such as the use of smoke,rustic greenhouses (Aguirre et al. 1999), and raisedbeds (Snchez de Lozada et al. 1998); and plantingfrost resistant potato cultivars. Since the latter methodis the most practical, a breeding program has beenestablished that aims to produce frost resistant potatocultivars similar to S. tuberosum (Carrasco et al.1997). Previously, successful breeding for frostresistance had been reported in the USA (Dearborn1969). This paper assesses the potential impact ofpotato cultivars with increased frost resistance inBolivia. This assessment is carried using a frameworkfor quantitative and constraint-specific agro-ecological zonation.

Agroecological zones (AEZs) stratify an area intoenvironmentally homogeneous domains usuallyderived from climate and soil data (e.g., FAO 1978-81, Kassam et al. 1991). AEZs are useful for selectingtest sites, interpreting experimental data, targetingtechnology, setting research priorities, and ex-anteimpact assessment (Wood and Pardey 1998). Forexample, Gryseels et al. (1992) used AEZs to setpriorities for the Consultative Group on InternationalAgricultural Research (CGIAR). AEZs can be classifiedas generic or specific, and as quantitative orqualitative. Qualitative AEZs can be ordinal or not.The International Potato Center (1991) uses a generic,qualitative zonation for global potato production.This zonation divides global potato production into 6zones ranging from "temperate" to "subtropicallowland". Generic zonation was also used in a FAO(1978-81) study in which production areas weredivided into four ordinal zones and designated "verysuitable" to "not suitable" for rainfed potato. Stol et al.(1991) and Van Keulen and Stol (1995) developed ageneric and quantitative zonation, using a simulationmodel and a climate and soil database within aGeographic Information System to calculate potentialand water-limited yield for global potato production.

The generic character of the AEZs describedabove makes them useful as a general reference thatcan be understood intuitively. However, they aredifficult to use when addressing specific researchquestions. Facilitated by the progress in informationtechnology and the development of geo-referenceddatabases, more flexible and specific approaches toagroecological zonation are emerging (Wood andPardey 1998, Corbett 1998). Hijmans (1999)


4

produced a specific zonation for frost-risk in potatoon the Altiplano in which interpolated, monthly,extreme minimumtemperature data were used tocalculate the probability of absolute minimumtemperature events throughout the year. Hijmans(1999) concluded that the introduction of potatocultivars with increased frost resistance couldstrongly reduce frost damage. If frost resistance inpotato increases from -1 to -2C, and damage onlyoccurs at -3C and lower, the percentage of thecurrent potato area with a frost event less than onceevery 10 yr nearly doubles, increasing from 18 to32%. However, this study did not quantify thepotential effect on yield of a potato cultivar withincreased frost resistance.

To estimate the potential impact of a newagricultural technology such as a frost resistantpotato cultivar, Wood and Pardey (1998) advocatethe use of specific agroecological zonation for theproblem at hand and then to estimate changes inyield or production costs for each zone. Estimatingyield loss due to frost damage is difficult because itdepends on the probability as well as the timing ofan absolute minimum temperature event. Forexample, a frost event of a specific magnitude in themiddle of the growing season will have a sharplydifferent effect on yield than one that occurs at theend of the growing season as most potatoes will havebeen formed by then. Eliciting estimates on frost-induced changes in yield from experts or farmers isone alternative (Valdivia et al. 1997), but may resultin highly subjective data. Producing a good estimatefor a large heterogeneous area can prove to be verydifficult.

This study demonstrates how crop growthsimulation models provide a useful alternative toeliciting techniques or classification criteria indetermining yield change from frost damage. Cropgrowth simulation models are mathematical descriptionof a crops response to the environment. Theyencapsulate our knowledge of eco-physiologicalprocesses, and they can be used to processenvironmental data as to produce easy-to-interpretoutput such as yield. By comparing the output ofdifferent model runs e.g., representing current andnew technology, the effect of technology adoptioncan be estimated. Instead of estimating differencesfor predefined agroecological zones, the model is runfor small grid cells and the results can be aggregatedby, e.g., administrative unit, or by production zone.Agroecological zones can be formed after thecalculations, using the model output.

MATERIAL AND METHODS

General framework

The general framework for constraint-specific andquantitative agroecological zonation is illustrated inFigure 2. As opposed to generic zonation approaches,the study is driven by a specific question. Theframework is flexible in the sense that the differenttypes of data used, and how they are combined,depends on the specific problem at hand as well asdata availability. Model choice depends on what isavailable and on the type and scale of correspondinginput data that is needed. In most cases, this includesdata on weather and soil. Once models are selected,different model runs can be compared using ancillarydata to interpret and aggregate output. Ancillary datawill typically include crop distribution and adminis-trative boundaries.

Simulation model

A slightly modified version of the LINTUL potatogrowth simulation model (Spitters, 1987; Stol et al.,1991) was used. The LINTUL model is based on athermal-time dependent description of ground cover.Ground cover is used to calculate interceptedradiation, and a constant radiation-use efficiency(RUE) parameter is used to calculate dry matterproduction. Allocation of dry matter to the variousorgans is thermal-time dependent. The effect of frostwas modeled using a simple damage function inwhich loss of ground cover is described as a linearfunction of minimum temperature between twotemperatures: the critical temperature (Tcr) and theleaves-dead temperature Tld (Figure 3). Above Tcrthere is no frost damage. If temperature drops belowTld, all leaves are lost. This type of relation wasdescribed by Sukumaran and Weiser (1972) forexcised leaflets of different potato cultivars andspecies. Subsequent ground cover expansion is alsoreduced by frost, depending on a linear functionbetween Tld and the regrowth temperature, Trg. Evenat Tld when the crop loses all its leaves, the crop maycontinue to grow. However, when the temperaturesdrop below Trg, growth ceases. We assume that frostdoes not have an effect on the radiation-useefficiency of the remaining foliage.

The simulation model was calibrated for the nativepotato cultivar 'Gendarme' using data collectedduring field trials in Patacamaya (La Paz Department,


5

Bolivia) in the 1998/99 season. In this experiment thepotato crop did not suffer any significant stress,although the growing season ended somewhatprematurely due to a number of frosts. Ground coverexpansion parameters were derived by fitting thecurves with biweekly observations. Radiation-useefficiency was calculated using observed groundcover and biomass from four sequential harvests. Theparameters for dry matter distribution betweenfoliage and tubers were also estimated from thesequential harvest data. The model was validatedwith data from Pomani (Department of La Paz,Bolivia) in 1996/97. In Pomani there was a mid-season, -2C frost on February 1, 1997 that led toconsiderable damage of crop foliage. Planting distance(0.7 x 0.3 m) and all other crop management practices,such as fertilization, were similar in both trials.

Weather data

A weather database described by INTECSA (1993)was used. The database consists of data from 139weather stations (Figure 1). For most weather stationsthere are at least 30 years of monthly records of total

precipitation and average minimum and maximumtemperature data. There is also a considerable amountof data on monthly extreme minimum temperature(44 stations), and other climate variables.

Monthly climate surfaces were generated forminimum and maximum temperature, absoluteminimum temperature and its standard deviation,and solar radiation. The ANUSPLIN programdeveloped by Hutchinson (1995, 1997) was used tointerpolate climate data from weather stations andproduce high-resolution climate surfaces (grids).ANUSPLIN fits Laplacian smoothing spline functionsof two or more independent variables (longitude,latitude, and usually elevation) through the climateobservations. This method relies on the strongdependence of climate (especially temperature) onelevation, but allows the size of this dependence tovary over time and space (Corbett, 1998). Elevationdata was taken from the U.S. Geological Survey'sGTOPO30 database, and used as an independent co-variable. The GTOPO30 and climate surface datahave a resolution of 30-arc seconds (approximately 1km2).

Figure 2. Flow diagram for quantitative and constraint-specific agroecological zonation using a simulation model and geo-referenced databases.

Question

Answer

Output

Model choice

&adaptation

Parameters

Queries

Input data

Ancillary data

Weather station data

Model

Experimental data

Information flow

Process

GIS Data

Data flow

Data

Question

Answer

OutputOutput

Model choice

&adaptation

Model choice

&adaptation

Model choice

&adaptation

ParametersParameters

QueriesQueries

Input dataInput data

Ancillary dataAncillary data



ModelModel

Experimental dataExperimental data

Information flow

Process

GIS Data

Data flow

Data


6

Figure 3. Description of frost damage (relative fraction ofground cover loss) as a function of minimum temperature for astandard S. tuberosum ssp. andigena potato cultivar from theAltiplano (Tcr = -1, Tld = -3). Estimated from Carrasco et al.(1997) and the Pomani experiment.

As the simulation model needs daily weather data,these were generated from the monthly climate data.For each grid cell 100 years of synthetic weatherwere generated to run the simulation model. Dailyminimum and maximum temperatures were calculatedby linear interpolation of the monthly averages. Onone random day of each month, minimumtemperature was simulated, using a random valuedrawn from a normal distribution described by themonthly average extreme minimum temperature andits standard deviation.

Potato distribution

A geo-referenced potato distribution database wascreated for the Altiplano. District level census data(INEI, 1993) was used for Peru, and departmentallevel census data (G-DRU, 1996) and maps of cropdistribution (ZONISIG, 1998) were used for Bolivia.

Estimating the effect of resistance

The potential yield of 'Gendarme' was comparedwith the potential yield of constructed genotypes thatdiffered from Gendarme in their level of frostresistance only. Frost resistance was described withfunctions like the one in Figure 3. The parametersused are presented in Table 1.

In the Altiplano, most potatoes are planted inOctober or November and harvested in April or May.For this study, we fixed emergence arbitrarily atNovember 25. Because of the stochasticity of theweather generator, the model was run 100 times per

grid cell, once for each generated year. The modeloutputs were averaged by grid cell and treatment.Then, the relative yield difference between theseaverages was calculated by grid cell. These resultswere weighted by the potato area in each grid celland tabulated.

Table 1. Different levels of frost resistance used in thesimulations. No damage occurs above critical temperature Tcr[C]. At Tld, 100% of the foliage is damaged. Whentemperatures drop below Trg, the crop stops growing. R-1 is thecurrent level of resistance in Solanum tuberosum ssp.andigena.

Resistance level Tcr Tld TrgR0 0 -2 -3R-1 -1 -3 -4R-2 -2 -4 -5R-3 -3 -5 -6R-4 -4 -6 -7R-5 -5 -7 -8

RESULTS

Model calibration

Based on the 1998 experiment, crop growthduration of 'Gendarme' (between emergence andsenescence) was estimated at 1250Cd (basetemperature = 0C), with tuberization starting at500Cd after emergence. Radiation-use efficiencywas estimated at 2.5 g/MJ (PAR), a value comparableto those reported in the literature (cf. Stol et al.,1991).

The simulation model somewhat overestimatesground cover and biomass production in the 1996experiment (Figures 4 and 5). This can partly beascribed to growth reduction due to excess water inthe experimental field. More data would be neededto further calibrate the model, but we considered itsufficiently accurate for use in this explorative study.

Effect of increased resistance

Changing frost resistance in Gendarme leads tosignificant changes in simulated potential yield. Thechange in simulated potential potato yield when frostresistance (Tcr) in 'Gendarme' is increased from -1 to-2C is shown in Figure 6. Only the area with thehighest 75% yield at current resistance levels is takeninto account. The other 25% is considered to be


7

predominantly planted with bitter potatoes. Thelikelihood of adoption of frost resistant cultivarswould be highest in areas with a high estimated yieldincrease. These areas would also be a good locationfor field trials with new genotypes.

Figure 4. Simulated (S) and observed (O) ground cover overtime for the 1996 and 1998 experiments. Other than frostresistance parameters, all data used to calibrate the model arefrom 1998. In the 1996 experiment, there was a 2C frost 58days after emergence, and in the 1998 experiment, there werefrosts of 2 to 4C 119, 120, and 121 days after emergence.

Figure 5. Simulated (S) and observed (O) total (tot) and tuber(tub) dry matter biomass production over time for the 1996 and1998 experiments. The model was calibrated with 1998 data,except for the frost resistance parameters. In the 1996experiment, there was a 2C frost 58 days after emergence,and in the 1998 experiment, there were frosts of 2 to 4C119, 120, and 121 days after emergence.

Yield response is especially strong when frostresistance changes from 0 to -3 C; after that theeffect levels off (Table 2; Figures 7 and 8), especiallywhen the area with a predominance of bitter potatoes

is eliminated. Increasing frost resistance from -1 to -2C over the whole potato area (excluding sectionswith a predominance of bitter potatoes) increasessimulated yield by 26%. For the Altiplano, this wouldamount to a yield increase from 6 to about 7.6 t/ha.The 1.6 tons/ha increase over 44,800 ha (eliminating30% of the total area, which is planted with bitterpotatoes) would lead to an average yearly increase inpotato production of 18%.

DISCUSSION AND CONCLUSION

Frost is an important constraint to potato productionin the Altiplano. We have demonstrated that theadoption of potato cultivars with increased frostresistance would lead to a strong decrease in yieldloss. Hence, breeding for increased frost resistanceseems to be a viable goal. Based on the datapresented here, we would suggest that Altiplanobreeding programs aim to develop cultivars with -2C frost resistance (i.e., a resistance increase of1C). This would have a major impact on yield (anaverage increase of 26%), and the probability ofsuccess in developing such cultivars seems high,given the relatively low increase in resistancerequired and the high levels of resistance available inwild and cultivated potatoes. After increasing currentlevels of resistance by more then 2C, the return ininvestment levels off, while the research cost willprobably increase.

This study only considered the current potatoarea. However, the introduction of new frost resistantpotatoes could lead to relative shifts (within the samearea) and absolute shifts (to other areas) in theAltiplano potato area. With the exception of theshores of Lake Titicaca, however, the Altiplano is notdensely populated and agriculture is not intensive, soland availability does not seem to be a limiting factorin potato production. Therefore, rather than shiftingproduction to new, colder areas, farmers wouldprobably choose to increase production in currentproduction zones. This would result in more efficientand less risky potato production. On the other hand,potential production areas at higher altitudes mightbe associated with better (less eroded) soil and moreprecipitation. Hence, some expansion of potato intohigher areas could be expected with the introductionof cultivars with increased frost resistance. However,more in-depth knowledge of farmer strategies inrelation to potato production and frost risk on theAltiplano is needed to predict farmers response.


8

Figure 6. The Altiplano (TDPS system) in the andes of Peru and Bolivia. Simulated yield increased for Gendarme when frostresistance increased from 1 to 2C.

Table 2. Simulated potential potato yield for Gendarme with different levels of imposed frost resistance (R0 is resistant (Tcr) to0C, R-1 to -1C, etc.). Averages calculated for the total Altiplano potato area and for the area excluding sections with apredominance of bitter potatoes (75% of the total potato area).

Resistance0 -1 -2 -3 -4 -5

All potato area (100%) 16.0 24.3 31.7 37.5 41.6 44.3Non-bitter potato area (75%) 20.6 30.6 38.6 43.7 46.6 48.0

This study focused on the potential impact ofbreeding potatoes with increased frost resistance butignored frost resistant species such as S. juzepzuckiiand S. ajanhuiri. With a few exceptions (Rea andVacher, 1992; Tapia and Saravia, 1997), thesespecies have not been the subject of scientificresearch. While they are often described as low

yielding, Tapia and Saravia (1997) reported yields of37 t/ha for S. juzepzuckii and 22 t/ha for S. ajanhuiri.These native species merit more basic and adaptiveresearch, and need to be considered when designingstrategies to diminish frost risk in potato productionon the Altiplano.


9

Figure 7. Cumulative distribution over total potato area in theAltiplano of simulated potential potato yields for Gendarmewith different imposed levels of frost resistance. R0 is resistantto 0C, R-1 to 1C, etc.

Figure 8. Simulated change in potential potato yield forGendarme with imposed different levels of frost resistance(Tcr= 0C to 5C), compared to that at the cultivars actuallevel resistance (Tcr= -1C). Averages calculated for totalpotato area in the Altiplano (100%), and for the area excludingsections with predominance of bitter potatoes (75%).

Although the simulation-based method used inthis study is more objective than other AEZapproaches, there are still numerous assumptionsand potential sources of systematic error. Systematicerror may have occurred because of the way theweather data was used but also because of otherreasons. Outcomes are sensitive to changes in modelparameters, notably in Tcr and Tld and in theparameters describing phenology.

Outcomes are highly dependent on the quality ofthe crop distribution data. This may very well be thegreatest source of error in our estimates. Cropdistribution data are important to obtain meaningful

results from GIS-linked models. National-level cropdistribution data are available through FAO, but dataat a lower level of aggregation are hard to get. Inorder to obtain more precise results in future AEZstudies, efforts to assemble crop distribution databasesshould be intensified.

Our method is an improvement over other studies,which used site-specific weather station data;assuming simulation results were representative ofeach respective surrounding area (e.g., Stol et al.,1991, Penning de Vries et al., 1996 and Waddel etal., 1999). A major disadvantage of that approach isthat the results become meaningless if weatherstations are far apart and comprise different climaticconditions, which is often the case in mountainregions. If climate data is interpolated before use in asimulation model, more meaningful results can beobtained.

A disadvantage of spatial interpolation of weatherdata is that the gain in spatial resolution generallycomes at the cost of temporal resolution. The choicebetween long-term, daily weather station data andinterpolated monthly average climate data hasimportant implications, since simulated yield frommonthly climate data (as per average weather patterns)does not necessarily coincide with the averagesimulation calculated from daily data (De Wit andVan Keulen, 1987; Nonhebel, 1994). We interpolatedmonthly average extreme temperature and itsstandard deviation to generate daily weather data. Acomparable but more elaborate approach, in whichboth average values and weather simulator parametersare interpolated, is described by Jones and Thornton(1999).

Although we used a very fine resolution grid, therecan still be important differences in microclimate thatwe have ignored. For example, frost is often worseon valley bottoms than on slopes, because ofnocturnal cold air drainage. In some areas, this isreflected by preferential use of hillsides, despite theirshallower soils. In the future, remotely sensedtemperature data might prove useful for improve thequality of the extreme minimum temperature surfaces(Franois et al., 2000).

Our method of creating daily minimum temperaturethrough linear interpolation with one randomextreme temperature per month is simple and couldbe improved. However, our study lacked time seriesof daily minimum temperature data, a prerequisitefor testing more elaborate approaches. This is animportant limitation because the results of this studyare strongly influenced by even small variation in


10

minimum temperature data. One should realize,however, that ex-ante studies couldnt always bepostponed long enough to allow for collection of allthe data required for optimal analysis. In this type ofstudy, researchers will have to strike a balancebetween ideal procedures and availability of dataand models (Figure 2).

With the various potential sources of error it ishard to speculate whether we would have over orunderestimated the potential for frost resistant potatoeson the Altiplano. But even if we overestimated theeffect somewhat, our main conclusions would notfundamentally change. Given the broad geneticvariability in frost resistance in wild and cultivatedpotato, and the strong simulated response in yield inthe R0 to R-3 classes, breeding for frost resistance inpotato for the Altiplano clearly has a high potentialimpact on potato yield.

REFERENCES

Aguirre G, Calderon J, Buitrago D, Iriarte V, Ramos J,Blajos J, Thiele G and Devaux A (1999) Rusticseed bed, a bridge between formal and traditionalpotato seed systems in Boliva. CIP Program Report1997-98. International Potato Center, Lima.

Booy A H (1961) Verband tussen nachtvorstschade aanaardappelen en bodemgesteldheid op Drentsedalgrond. Landbouwvoorlichting (januari 1961):10-18.

Canahua A, and Aguilar P C (1992) Agroecologica de laspapas amargas en Puno. In: Rea and Vacher,pp.57-61.

Carrasco E, Devaux A, Garca W and Esprella R (1997)Frost tolerant potato varieties for the AndeanHighlands. In: Program Report 1995-1996.International Potato Center, Lima. pp. 27-232.

Corbett J D (1998) Classifying maize production zones inKenya through multivariate cluster analysis. In:Hassan, R.M. (ed.), Maize technology developmentand transfer. A GIS application for research andplanning in Kenya. pp.15-25.

Dearborn C H (1969) Alaska Frostless: an inherently frostresistant potato variety. American Potato Journal46: 1-4.

De Bouet du Portal D (1993) Etude des gelees surl'Altiplano Bolivien. ORSTOM and ENSAM,Montpellier. p.41.

De Wit C T and van Keulen H (1987) Modeling theproduction of field crops and its requirements.Geoderma 40: 253-265.

FAO (Food and Agricultural Organization of the UnitedNations), (1978-81) Report on the Agro-ecologicalzones project. World soils resources report 48.Vol 1-4. FAO, Rome.

Franois C, Bosseno R, Vacher J J and Seguin B (2000)Frost risk mapping derived from satellite andsurface data over Bolivian Altiplano. Agricultureand Forest Meteorology. In press.

Frre M, Rijks J Q and Rea J (1975) Estudio agroclimatolgicode la zona Andina. FAO, Rome, p.375.

G-DRU (1996) (Grupo Interinstitucional de DesarrolloRural), Anuario Estadstico del Sector rural 1995-1996 de Bolivia. Centro de Informacin para elDesarrollo, La Paz. p.297.

Gryseels G, De Wit C T, McCalla A, Monyo J, Kassam A,Crasswell E and Collinson M (1992) Settingagricultural research priorities for the CGIAR.Agricultural Systems 40, 59-103.

Hijmans R J (1999) Estimating frost risk in potatoproduction on the Altiplano using interpolatedclimate data. Impact on a changing world.Program Report 1997-1998. International PotatoCenter, Lima. pp.373-380.

Huanco S (1992) Potencial de las papas amargas en elAltiplano de Puno. In: Rea and Vacher, pp.25-26.

Hutchinson M F (1995) Interpolating mean rainfall usingthin plate smoothing splines. Int. J. GeographicalInformation Systems 9(4): 385-403.

Hutchinson M F (1997) ANUSPLIN version 3.2. UserGuide. Australian National University. Center forResource and Environmental Studies. Canberra.p.39.

INEI (1993) Tercer Censo Nacional Agropecuario.Database. INEI, Lima.

INTECSA (1993) Estudio de Climatologa. Plan directorglobal binacional de proteccin prevencin deinundaciones y aprovechamiento de los recursosdel lago Titicaca, ro Desaguadero, lago Poopo ylago Salar de Coipasa (Sistema T.D.P.S).INTECSA, AIC, CNR. p.85.

Kalma J D, Laughlin G P, Caprio J M and Hamer P J C(1992) The bioclimatology of frost. Advances inBioclimatology 2. Springer-Verlag, Berlin. p.144.

Knapp G (1988) The effects of variations in meantemperature and frost risk. In: Parry M L, Carter TR and Konijn NT. The impact of climaticvariations on agriculture Volume 2: assessmentsin semi-arid regions. Part IV, The impact ofclimatic variations on agriculture in the Sierra ofEcuador. Kluwer, Dordrecht. pp. 443-459.

Le Tacon, Ph (1989) Manifestation des risques climatiquesa l'chelle de l'exploitation agricole, consequenessur les pratiques paysannes. Cas de l'Altiplano


11

Bolivien. Memoire de stage. ORSTOM, ENSSAA,CNEARC, ENSAM. p.130.

Morlon P (1989) Du climat la commercialisation:lexample de lAltiplano pruvien. In: Eldin M,and Milleville P. Le risque en agriculture.ORSTOM, Paris. pp.187-224.

Nonhebel S (1994) The effects of use of average instead ofdaily weather data in crop growth simulationmodels. Agricultural Systems 44: 377-396.

OEA (1996) Diagnstico ambiental del sistema Titicaca-Desaguadero-Poopo-Salar de Coipasa (SistemaTDPS), Bolivia Peru. Organizacin de los Esta-dos Americanos, Washington. p.192.

Rea J (1992) Vigencia de las papas nativas en Bolivia. In:Rea and Vacher. p.15-23.

Rea J and Vacher J J (1992) La papa amarga. Primera mesaredonda: Per - Bolivia, La Paz 7-8 Mayo 1991.ORSTOM, La Paz. p.111.

Snchez de Lozada D, Baveye P and Riha S (1998) Heatand moisture dynamics in raised field systems ofthe lake Titicaca region (Bolivia). Agricultural andForest Meteorology 92(4): 251-265.

Smith C T, Denevan W M and Hamilton P, (1968) Ancientridged fields in the region of Lake Titicaca.Geographical Journal 134: 353-367.

Sukumaran N P and Weiser C J (1972) An excised leaflettest for evaluating potato frost tolerance.Hortscience 7(5): 467-468.

Tapia N and Saravia G, (1997) Biodiversidad en papasamargas. Prov. Tapacar, Dpto. de Cochabamba.AGRUCO, Cochabamba, p. 53.

Terrazas F, Suarez V, Gardner G, Thiele G, Devaux A andWalker T (1998) Diagnosing potato productivityin farmers' fields in Bolivia. Social ScienceDepartment Working Paper 1998-5. InternationalPotato Center, Lima, Peru.

Stol W, De Koning G H J, Kooman P L, Haverkort A J, VanKeulen H, Penning de Vries F W T (1991) Agro-ecological characterization for potato production.A simulation study at the request of theInternational Potato Center (CIP), Lima, Peru.CABO-DLO, Report 155. p. 53.

Valdivia-A R, Quiroz R, Valdivia-F R and Choque-huancaV (1997) Risk Analysis of Potato Production in theAltiplano: Quantifying Farmers Beliefs. CIPProgram Report 1995-96. International PotatoCenter, Lima. pp. 29-33.

Van Keulen, H. and Stol W, (1995) Agro-ecologicalzonation for potato production. In: Haverkort A J,and MacKerron D K L (eds.), Potato Ecology andModeling of Crops under Conditions LimitingGrowth. Kluwer, Dordrecht. pp. 357-379.

Wood S and Pardey P G (1998) Agroecological aspects ofevaluating agricultural R&D. Agricultural Systems57(1): 13-41.

ZONISIG (1998) Zonificacin agroecolgica y socio-econmica de la Cuenca del Altiplano delDepartamento de La Paz. Ministerio de DesarrolloSostenible y Planificacin, Prefectura delDepartamento de La Paz. DHV-ITC, La Paz.p.213.

INTRODUCTIONMATERIAL AND METHODSGeneral frameworkSimulation modelWeather dataPotato distributionEstimating the effect of resistance

RESULTSModel calibrationEffect of increased resistance

DISCUSSION AND CONCLUSIONREFERENCES

cartofi altiplano.2

Documents